Methodology
How the snapshots are taken, on which dates, which pages are left out and why, and what the data cannot tell you.
Page written on 2026-10-05. It describes the snapshot of 2026-10-05.
AI-assisted, compiled by automated agents. SaaS Pricing Changes is compiled by automated agents from snapshots of public pricing pages. A language model reads each page and extracts the plans and prices; a script compares the snapshots and writes the issue. No person reads the pages or the issue before it is published. Prices may have changed since the snapshot date: confirm on the vendor's own pricing page before you act on a figure.
What is tracked
60 SaaS tools, one public pricing page each. By category: developer tools 12, analytics 8, email 7, hosting 6, support 6, CRM 5, design 5, project management 5, payments 3, productivity 3.
For each tool the snapshot records facts only: plan names, listed prices, the billing unit (per month, per month billed annually, per year, per seat, usage rate), whether a free tier exists, numeric limits as the page states them, and whether a plan says "contact sales". No page text is stored or republished.
Which pages are fetched: the terms and robots.txt check
Before any pricing page entered the tracker, three documents were read for each candidate tool on 2026-10-05:
- The robots.txt file of the host that serves the pricing page, for the
User-agent: *group. The pricing path must not match a Disallow rule. - The terms that govern the public website (the website terms of use when the company publishes them, otherwise the general terms of service), searched for clauses on robots, spiders, scrapers, crawlers, data mining and automated access.
- The pricing page itself, to confirm that prices appear in the page text without a login.
144 candidates were examined to reach 60 tracked tools:
| Outcome | Tools | What it means |
|---|---|---|
| Tracked | 60 | robots.txt does not disallow the pricing path and no clause against automated access was found in the terms read. |
| Excluded: forbidden | 41 | The terms contain a clause against automated access, robots, crawling or scraping. A clause that names only the product, not the website, is still counted as forbidden. |
| Excluded: unverified | 36 | The terms or robots.txt could not be read, the fetched page held no prices, or the fetch was blocked. These are left out, not guessed. |
| Held back | 7 | Passed every check made, but on thinner evidence than the tracked tools. |
What "tracked" does and does not mean:
- All 60 tracked tools are classified as silent: no restricting clause was found. None of the terms read gives express permission to crawl, so silence is not presented here as permission.
- Many well-known tools are absent because their terms forbid automated access. Their absence says nothing about their prices.
- robots.txt was read for
User-agent: *only, and "silent" refers to the one terms document read per tool. A linked policy that was not read could contain a restriction. - Terms change. Each tool carries the date it was checked, and a check older than 90 days is repeated before the tool stays in the tracker.
How a snapshot is taken
- One request per pricing URL, with no retries and no query parameters. A weekly fetch of a single public page per tool.
- The page is converted to text and a language model extracts the facts, with the instruction to report only what the page literally states.
- Nothing is computed: where a page gives only a discount percentage, the discounted price is left empty.
- A free trial is not counted as a free tier.
- A promotion (for example a temporary discount for new customers) is kept as a note and does not replace the list price.
- Pages that list many products (cloud hosts, usage-priced APIs) are tracked through a fixed set of headline items per tool, so that each comparison looks at the same items.
Dates
| Snapshot | Source | Tools read |
|---|---|---|
| 2026-10-05 | Live pricing pages | 60 of 60 |
| Target 2026-07-01 | Archived captures in the Internet Archive Wayback Machine | 0 of 60 |
The plan for issue 1 was to compare the live pages with archived captures from about 2026-07-01. The Wayback Machine's availability index was queried once per tool and returned a capture within 31 days of that date for 51 tools, a capture from 2026-09-09 for one, a capture from 2021 for one, and no capture for seven. The content of those captures could not be read by the tooling used for this build, so no earlier price was recorded for any tool. No figure was filled in from memory or estimated.
As a result, issue 1 reports no changes. It is a baseline: the plans and prices listed on the 60 pages on 2026-10-05. Changes are reported from the next snapshot onward, each with the two dates being compared. Snapshots are planned weekly; every issue states the dates of the snapshots it actually used.
How changes are detected
A script compares two snapshots of the same tool and reports seven kinds of change: price increase, price decrease, plan added, plan removed, billing unit changed, free tier added, free tier removed.
- A tool is compared only when it was readable in both snapshots. Tools skipped for an issue are listed in that issue with the reason.
- Prices are compared only within the same currency and the same billing unit. A page served in another currency is not a price change.
- A figure whose billing period the page does not settle (for example one number next to a monthly/annual toggle) is recorded as a displayed figure and is never compared with a monthly price.
- The comparison script is tested against 12 fixtures (price up, price down, plan added, plan removed, billing unit changed, free tier added and removed, unreadable entries skipped, no false changes, malformed input refused, low-confidence reads flagged, disclosures present in the issue). All 12 passed on the run that produced issue 1.
Limits of the data
- Single read, single source. Each page was read once by an automated agent and nothing was checked against a second source. An extraction error is possible on any entry.
- Low-confidence reads. 14 of the 60 entries in the 2026-10-05 snapshot are marked "low": HubSpot, Zoho CRM, Buttondown, MailerLite, Mailjet, Honeybadger, Cloudflare, n8n, Help Scout, Front, Stripe, monday.com, Shortcut and Calendly. The usual reason is a monthly/annual toggle or a slider that the page text does not resolve. A change on one of these tools is flagged as low confidence in the issue.
- Location. Pages are served for the location of the fetch. In the 2026-10-05 snapshot Zoho CRM came in INR, and Pitch and n8n in EUR. Your region may see other prices.
- Currency symbol. Where a page shows a "$" sign without naming the currency, it is recorded as USD and the entry says so.
- Partial views. Some entries cover one product line only: HubSpot is the Marketing Hub view, monday.com is Work Management, and the Cloudflare entry holds developer-platform usage prices, not the application plans.
- List prices only. Negotiated, enterprise and reseller prices are not visible on a public page. Taxes are not included unless the page includes them.
- Weekly resolution. A price that changes and reverts between two snapshots is not seen. The date of a change is the date of the snapshot that first shows it, not the date the vendor made it.
- Not advice. The tracker reports what pages listed on given dates. It is not a recommendation to buy, switch or cancel anything.
Corrections
When a later snapshot shows that an earlier entry was misread, the next issue says so in its own section, with the tool, the wrong figure and the corrected one. Past issues are not silently rewritten.